Chapter 7. Functional role of CSTA in breast cancer
7.2. Results
6.2.5. Upstream CpGs methylation inversely correlates with CSTA expression in breast tumors
CSTA expression (RNA-Seq; log2(RPKM+1)) and methylation data (generated with Illumina Infinium® Human Methylation 450K BeadChip array) for the primary breast tumors in the TCGA breast cancer dataset were accessed using the UCSC Xena browser266. Methylation data were available for 4 probes in the CSTA locus. These probes correspond to the four CpG sites in Region 1 shown in Figure 6.3A. Firstly, probe-wise analysis of the correlation between methylation (beta value) and CSTA expression in primary tumors was performed. Methylation at each of the 4 CpG sites inversely correlated with CSTA expression (Table 6.1). The primary tumors were divided into two groups, namely hypo- and hyper- methylated, using a beta value of 0.3 as a cut-off. CSTA expression in hypo-methylated tumors was significantly higher than those in hyper-methylated tumors (Figure 6.4A-D). A composite methylation score for each sample was generated by averaging the beta values for all the
probes. The composite methylation score correlated inversely with CSTA expression (ρ = -0.582, p < 0.0001, Figure 6.4E). Furthermore, when the tumors were divided into hypo-
and hyper-methylated groups, based on a cut-off composite methylation score of 0.3, the hypo-methylated tumors showed significantly higher expression of CSTA compared to hyper-methylated tumors (Figure 6.4F).
Table 6.1. Correlation between CSTA expression and methylation in CSTA locus in breast tumors of the TCGA cohort.
6.2.6. Intron-2 region of CSTA encompasses a potential ERE
In chapter 5, estrogen-mediated suppression of CSTA, which occurs via binding of ERα to intron-2 region of CSTA, was demonstrated. Analysis of the CSTA locus using JASPAR294 revealed a potential ERE in the intron-2 region. (Figure 5.5 and Figure 5.6). Interestingly, this ERE was located amidst the differentially methylated CpG sites analyzed in Region 2 (Figure 6.5).
Probe ID Spearman’s ρ p-value
cg14664412 -0.5482637 < 0.0001 cg18618429 -0.5683092 < 0.0001 cg21932814 -0.4717445 < 0.0001 cg26928972 -0.5154166 < 0.0001
Figure 6.4.Inverse correlation between CpG methylation and CSTA expression in breast tumors of the TCGA cohort. A-D. Probe-wise analysis of the correlation between methylation and CSTA expression. The tumors were segregated into hypo-methylated or hyper-methylated groups based on the threshold beta value of 0.3 for each probe. The distribution of CSTA expression in hypo-methylated and hyper-methylated tumors are represented as box plots. E-F. A composite methylation score, which is the average beta value of all the probes, was determined for each tumor sample. The scatter plot of the composite methylation score versus CSTA expression is shown in E. The tumors were segregated into hypo-methylated or hyper-methylated groups based on the threshold composite score of 0.3. The distribution of CSTA expression (log2(RPKM+1)) in hypo- methylated and hyper-methylated tumors is represented as a box plot (F). The difference in mean CSTA expression in hypo-methylated and hyper-methylated tumors was analyzed by Welch two-sample t-tests.
A B
C D
F
cg18618429 (p < 0.0001)
cg21932814 (p < 0.0001) cg14664412 (p < 0.0001)
Composite (p < 0.0001) cg26928972 (p < 0.0001)
CSTA expression
Hypo-methylated Hyper-methylated 10
8 6 4 2 12
CSTA expression
Hypo-methylated Hyper-methylated 10
8 6 4 2 12
CSTA expression
Hypo-methylated Hyper-methylated 10
8 6 4 2 12
CSTA expression
Hypo-methylated Hyper-methylated 10
8 6 4 2 12
CSTA expression
Hypo-methylated Hyper-methylated 10
8 6 4 2 12
ρ = -0.582, p < 0.0001
Composite methylation score
CSTA expression 0.6
0.4
0.2 0.8
2 4 6 8 10 12
E
The effect of global demethylation on estrogen regulation of CSTA was studied in MDA-MB-231 and T47D cells. ERα and CSTA expression in 5-aza-untreated or -pretreated MDA-MB-231 and T47D cells, which were stimulated with vehicle or 10 nM E2 was examined. 5-aza caused a significant loss of methylation in Region 2 (p = 0.041 in MDA-MB- 231, p = 0.034 in T47D) (Figure 6.6). As expected, in 5-aza-untreated MDA-MB-231 cells, ERα protein expression was not detectable after vehicle or E2 treatment (Figure 6.7A, lanes 1 and 2). There was no significant difference in CSTA mRNA expression (Figure 6.7C, bars 1 and 2, ANOVA followed by Tukey’s HSD). On the other hand, in 5-aza-pretreated cells, an immunoreactive protein was detected on western blots with ERα-specific antibody (Figure 6.7A, lanes 3 and 4). This immunoreactive protein had a higher molecular mass compared to the expected 66 kDa for ERα. Notwithstanding this discrepancy, induction of PR, and further enhancement of its expression with E2 confirmed the generation of a functional ERα in 5-aza pretreated cells (Figure 6.7A, B, lanes 3 and 4). 5-aza significantly induced CSTA mRNA expression in MDA-MB-231 cells (Figure 6.7C, bars 1 and 3; ANOVA followed by Tukey’s HSD). E2 suppressed the 5-aza induced levels of CSTA mRNA, although the difference was not statistically significant when analyzed by ANOVA. However, the levels of CSTA mRNA in 5-aza pretreated cells with and without E2 treatment were significantly
different when analyzed by the Welch two-sample t-test in (Figure 6.7C, bars 3 and 4, p = 0.0098). Western blots failed to demonstrate CSTA protein in MDA-MB-231 cells. E2
treatment resulted in ERα occupancy in intron-2 in 5-aza pretreated cells (Figure 6.7D, lanes 8 and 9). These results show that demethylation of intron-2 CpGs leads to restoration of ERα and CSTA expression and estrogen suppression of CSTA in MDA-MB-231 cells.
T47D cells express a very low or undetectable level of CSTA. Without 5-aza pretreatment, T47D cells treated with E2 showed decreased levels of ERα protein and
B A
TTCTGCTATCAAACTTTTCCTACTGGATCTCAGCCACCGATCCCAGTTCCCTTTTACTTC CTGGTAGTCTGGCTGTTGATCCCTTTGCTCTGAGGCACTCTAGATTTAAGGTCTTGCCAG TGATGTGACCTTCTCTATGTATTTCAAGTACCTATCAAGAGGTAGGTGGTAGAATGGAAG GACCACAAGCTTAGGTGTCAGAGTGTCCTGGGTTTGAACCCTTGTTCAATTTGTTCTATG GGAAGCTCCTCCTCCTCTCTGAGCCTTCATTCCCTTATCTGCACAATGAGGGTAATAATC TACTTCGCAGCGTGTTGTGAGGAATAAATAAGCTGGAAATTTATTGAGCACTTATAATTC ACTATGCACTATTCTAAGAACAGGGCTT
increased levels of PR, as expected (Figure 6.8A, B, lanes 1 and 2)303,304. There was no observable effect on CSTA protein (Figure 6.8A, lanes 1 and 2). However, an increase in CSTA mRNA was observed, although the increase was not statistically significant (Figure 6.8C, bars 1 and 2; ANOVA followed by Tukey’s HSD). 5-aza pretreatment alone caused a decrease in ERα protein in T47D cells (Figure 6.8A, lanes 1 and 3) in a manner similar to that reported in MCF-7 cells305. This also led to increased CSTA (Figure 6.8A, lanes 1 and 3) and decreased PR protein expression (Figure 6.8B, lanes 1 and 3). E2 induction of PR in 5-aza pretreated T47D cells showed that ERα was functional (Figure 6.8B, lanes 3 and 4). E2 not only enhanced CSTA protein expression (Figure 6.8A, lanes 3 and 4) but also significantly enhanced CSTA mRNA in 5-aza pretreated cells (Figure 6.8C, bars 1, 3 and 4;
ANOVA followed by Tukey’s HSD). E2 treatment resulted in ERα occupancy in the intron- 2 region in 5-aza treated cells (Figure 6.8D, lanes 9 and 10). These results show that demethylation of intron-2 CpGs restores estrogen regulation of CSTA in T47D cells.
Figure 6.6.5-aza treatment demethylates Region 2 in MDA-MB-231 and T47D cells. Cells were treated with 10 µM 5-aza for 5 days. gDNA isolated from treated cells were bisulfite converted and used for PCR reactions with Region 2-specific primers. The PCR amplified products were cloned in TA vector and sequenced.
13 and 15 independent TA clones were analyzed for methylated and unmethylated CpG sites in Region 2 of MDA-MB-231 and T47D cells, respectively. The methylation pattern is represented by lollipop plots. Filled circles represent methylated CpGs, and open circles represent unmethylated CpGs. The proportion of methylated CpGs are indicated in parentheses. The proportions were tested for significant difference as described in materials and methods. p-values obtained from Welch two-sample t-test are indicated above the plots.
MDA-MB-231 Control
(0.59)
MDA-MB-231 5-aza (0.36)
* p = 0.041
A T47D
Control (0.60)
T47D 5-aza (0.38)
* p = 0.034
B
76 Results
Figure 6.7.Global demethylation restores estrogen regulation of CSTA in MDA-MB-231 cells. A-C. Cells were subjected to global demethylation using 10 µM 5-aza for 5 days. The cells were then stimulated with 10 nM E2 or ethanol (vehicle) for 24 h. ER and PR expression was analyzed by western blotting (A, B) and CSTA expression was analyzed by semi-quantitative RT-PCR (C). CycA was used as an internal control in semi- quantitative RT-PCR. Histone H3 served as an internal control in western blots. CSTA mRNA expression in the control samples (without 5-aza and E2 treatments) were set to 1 and the expression in the other treatment groups was expressed relative to the control. Bars represent mean relative expression ± S.D. Data were analyzed by ANOVA followed by Tukey’s HSD (n = 3). *p < 0.05. D. Cross-linked chromatin samples from the treated and control cells were fragmented and immunoprecipitated with monoclonal ERα- or IgG-specific antibodies.
Immunoprecipitated DNA was reverse cross-linked, purified and subjected to PCR analysis using primers specific for Region 2 or pS2. Note the enrichment of the ERE containing sequence in the pS2 locus following E2 treatment (with 5-aza pretreatment), which validated the ChIP protocol.
15 20 100 (kDa)
ERα
Histone H3 A
Lane 1 2 3 4
75 100 150 (kDa)
25
PR
Histone H3 B
1 2 3 4 Lane
0 0.5 1 1.5 2 2.5 3
Relative expression
CSTA CycA C *
CSTA-Region 2 pS2 5-aza 10 nM E2
Input ERα
Water control
IgG
- + - + - + - + - + - + - - + + - - + + - - + +
Lane 1 2 3 4 5 6 7 8 9 10 11 12 13 14 D
5-aza 10 nM E2
- -
- -
+ + + +
5-aza 10 nM E2
- -
- -
+ + + +
5-aza 10 nM E2
- -
- -
+ + + +
Figure 6.8. Global demethylation restores estrogen regulation of CSTA in T47D cells. A-C. Cells were subjected to global demethylation using 10 µM 5-aza for 5 days. The cells were then stimulated with 10 nM E2 or ethanol (vehicle) for 24 h. CSTA, ER and PR protein expression was analyzed by western blotting (A, B).
CSTA mRNA expression was analyzed by qRT-PCR (C). CycA was used as an internal control in qRT-PCR.
Histone H3 served as an internal control in western blots. Relative CSTA expression in the control samples (without 5-aza and E2 treatments) were set to 1 and the expression in the other treatment groups was expressed relative to the control. Bars represent mean relative expression ± S.D. Data were analyzed by ANOVA followed by Tukey’s HSD (n = 4). ** p < 0.01, ***p < 0.001. D. Cross-linked chromatin samples from the treated and control cells were fragmented and immunoprecipitated with monoclonal ERα- or IgG-specific antibodies.
Immunoprecipitated DNA was reverse cross-linked, purified and subjected to PCR analysis using primers specific for Region 2 or pS2. Note the enrichment of the ERE containing sequence in the pS2 locus following E2 treatment, which validated the ChIP protocol.
Histone H3
ERα
Histone H3
0 2 4 6 8 10 12
C
α
- + - +
Lane 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
5-aza
- -
78 Discussion
Similar experiments in MCF-7 cells showed that global demethylation neither affected CSTA expression nor affected E2-mediated suppression (Figure 6.9). Without 5-aza pretreatment, MCF-7 cells treated with E2 showed a significant reduction in CSTA mRNA (Figure 6.9A, bars 1 and 2) and significant induction in pS2 mRNA (Figure 6.9B, bars 1 and 2). In 5-aza pretreated cells, E2 significantly reduced CSTA mRNA (Figure 6.9A, bars 3 and 4) and induced pS2 mRNA (Figure 6.9B bars 3 and 4). No significant difference in CSTA (Figure 6.9A, bars 2 and 4) and pS2 expression (Figure 6.9B, bars 2 and 4) was observed in E2-treated cells with or without 5-aza pretreatment.
Figure 6.9. E2-mediated suppression of CSTA is unaffected by global demethylation in MCF-7 cells.
A, B. MCF-7 cells were subjected to global demethylation using 10 µM 5-aza for 5 days. The cells were then stimulated with 10 nM E2 or ethanol (vehicle) for 24 h. CSTA and pS2 expression were analyzed by qRT-PCR.
CycA was used as an internal control. pS2 was used as positive control for E2 treatment. aRelative CSTA and pS2 mRNA expression data are represented as bar graphs. The expression in the control samples (without 5-aza and E2 treatments) was set to 1 and the expression in the other treatment groups was expressed relative to the control. Bars represent mean relative expression ± S.D. Data were analyzed by ANOVA followed by Tukey’s HSD (n = 4). ** p < 0.01, ***p < 0.001.
the inverse relationship between CSTA expression and methylation in the context of breast cancer. The CSTA locus lacks CpG islands. However, the examples of DNA methylation- dependent regulation of CpG island-less genes310-316 motivated the investigation of DNA methylation of CSTA locus.
The conclusions of this study are drawn from results obtained through three different approaches; a) analysis of CSTA expression and methylation data from the TCGA breast cancer cohort, b) examination of DNA methylation data (ENCODE project) for the CSTA locus, and c) bisulfite sequencing of DNA isolated from breast cancer cell lines, which express differential levels of CSTA. TCGA methylation data was generated using the Infinium Methylation 450K BeadChip arrays, which do not have probes to interrogate all CpG sites in the CSTA locus. Due to this limitation, no conclusion could be drawn regarding the correlation between CSTA expression and methylation of the intron-2 CpGs. Nevertheless, an inverse correlation between CSTA expression and methylation in the upstream CpG sites was observed (Figure 6.4E). The ENCODE project data corresponding to MCF-7 and T47D cells, and the bisulfite sequencing results clearly demonstrated that CSTA expression, and methylation in the intron-2 CpGs, are inversely correlated. Collectively, these data provide compelling evidences in favor of DNA methylation-dependent silencing of CSTA in breast cancer cells. Altered cathepsin B: CSTA ratio in breast tumors is reported. It also correlates with disease prognosis30,32,226. Ablation or inhibition of cathepsin B also inhibits spine and lung metastasis in the animal model35. Therefore, it was proposed that DNA methylation-mediated silencing of CSTA in primary breast tumors tips the cathepsin B/CSTA balance in favor of cathepsin B, which in turn facilitates tumor invasion and metastasis. A detailed study on the correlation of DNA methylation in the CSTA locus, and disease progression, treatment outcome and survival, may uncover its potential as a prognostic marker.
CSTA and ERα mRNA expression in breast tumors of the TCGA cohort are inversely correlated (Figure 4.6A). This is consistent with the observed induction of CSTA mRNA in MCF-7 cells following ERα knockdown (Figure 5.4B). The inverse relationship is reiterated in the ERα-positive breast cancer cell lines used in this study. ZR-75-1, which has the highest expression of CSTA, has the least ERα expression, whereas T47D, which has the least expression of CSTA, has the highest expression of ERα. CSTA and ERα expression levels in MCF-7 cells are in between these two extremes. Estrogen suppresses CSTA expression in MCF-7 cells via ERα (Figure 5.5 and Figure 5.6). However, the mechanistic role of ERα was not known. The prediction of an ERE by JASPAR, the peak of ERα binding revealed by ChIPseq data, and the validation of increased ERα occupancy in the intron-2 following
80 Discussion
estrogen treatment of MCF-7 cells suggest that estrogen suppresses CSTA expression at the level of transcription. The precise events post-ERα binding that lead to transcriptional shut- off are worth addressing in future investigations. However, the mechanism of estrogen- mediated regulation of CSTA is more complex. Estrogen does not produce similar effects on CSTA expression in ERα-positive breast cancer cell lines. Estrogen suppresses CSTA expression in ZR-75-1 cells. However, the extent of suppression is much lower than that observed in MCF-7 cells. In T47D cells, estrogen does not modulate CSTA expression.
Subsequent analysis revealed the probable reason behind the lack of estrogen-mediated regulation of CSTA in T47D cells.
Here two specific regions, namely Region 1 and Region 2, that encompass few of the upstream and intron-2 CpG sites, respectively, were analyzed. While it is worth analyzing methylation at every CpG site in the CSTA locus, this study shows that intron-2 is the site of convergence of estrogen regulation and DNA methylation-dependent silencing. Interestingly, the ERα binding site is located amidst the intron-2 CpGs (Region 2). The global demethylation experiments with MDA-MB-231 and T47D cells revealed the conflict between ERα binding and DNA methylation in Region 2. Due to methylation-dependent silencing, MDA-MB-231 cells do not express ERα317,318 and CSTA (Figure 6.1A, B). Global demethylation in MDA-MB-231 cells established functional ERα (as revealed by PR induction upon estrogen stimulation), and CSTA mRNA expression (Figure 6.7A, B). Furthermore, estrogen tends to suppress 5-aza induced CSTA mRNA, resembling estrogen regulation of CSTA in MCF-7 cells. This was possible because demethylation of Region 2 CpGs made the intron-2 ERE accessible to ERα. The conflict is also supported by the results from T47D cells. It must be noted that T47D cells show a significantly greater level of Region 2 methylation than MCF-7 cells, which likely prevents ERα binding to the ERE. This is arguably the reason why despite detectable levels of CSTA and functional ERα in T47D, estrogen does not regulate CSTA. 5-aza not only increased CSTA expression in T47D but also made it amenable to estrogen regulation. Although it is not clear why the direction of CSTA regulation in T47D is opposite to that observed in MCF-7, MDA-MB-231 and ZR-75-1 cells, these are enticing evidence that indicates the crucial role of intron-2 in CSTA expression and regulation.
The relationship between DNA methylation and transcription is not a one-way interaction. In a given genomic locus, the transcriptional activity can limit DNA methylation.
In Arabidopsis, this is evident from the distribution of methylated and transcriptionally active loci319. Pharmacological inhibition of RNA polymerase II induces repressive histone modification, which results in epigenetic silencing. It is possible that repressive histone
modification also leads to DNA methylation. Thurman and co-workers studied methylation of transcription factor binding sites and transcription factor abundance in DNAse I hypersensitive sites. They found an inverse correlation between the expression level of a given transcription factor and methylation of the cognate binding site. This suggests a model of
“passive DNA methylation”320. On the other hand, methylation of cytosine residues in CpG dinucleotides prevents binding of transcription factors to their cognate response elements on DNA321-324, thereby interfering with gene expression. An exception to this model is the binding of Sp1 to the methylated cognate site, leading to enhanced gene transcription325. Alternatively, methylated CpG sites attract MCBPs326. MCBPs, in turn, recruit histone deacetylases and methylases that cause remodeling and compaction of the local chromatin and transcriptional shutoff. This is the converse model of methylation-mediated blockade of transcription factor access. In the context of the interaction between transcriptional activity and transcription factor binding, ERα is not an exception. Ung and co-workers have analyzed DNA methylation in relation to ERα expression and binding327. They found an inverse correlation between ERα expression and CpG methylation within ERα binding sites. Methylation of these CpGs was therefore interpreted as being dependent on ERα activity, consistent with the passive model of DNA methylation; more the ERα binding, lesser the methylation. Except for T47D, the cell lines used in this study have patterns of ERα expression and Region 2 methylation that is consistent with the passive model. However, that methylation of Region 2 in MDA-MB-231 and MDA-MB-453 cells is entirely due to lack of ERα expression, cannot be stated with certainty. Parallelly, results from T47D cells are consistent with the converse model, wherein despite ERα expression and detectable levels of CSTA, estrogen could not significantly modulate CSTA expression. The combined results from MDA-MB-231 and T47D showed that demethylation could restore estrogen regulation at the CSTA locus. This is most likely due to the restoration of the ERα access to the ERE.
Taken together, the present study shows that CSTA expression in breast cancer cells is inversely related to DNA methylation in the CSTA locus. It explains the loss of CSTA expression in breast tumors. DNA methylation of CSTA may be exploited for predicting metastatic progression of breast tumors. Furthermore, this study has uncovered an interesting interplay between ERα binding and transcriptional regulation. The proposed model from this study is that CSTA expression in breast cancer cells is an integrated result of estrogen regulation and DNA methylation-dependent silencing converging on the intron-2.
7.1. Introduction
Metastasis accounts for approximately 90% of cancer-related deaths328. Metastasis is the major hurdle in the treatment of breast cancer, necessitating therapeutic intervention to specifically target metastasis. A better understanding of molecular mechanisms of tumor progression may unravel the key molecules against which therapeutic strategies could be designed. The basic strategies of intervention are a) blocking the molecules which promote
metastasis using specific antibodies, b) restoring the expression of suppressors, or c) mimicking the function of suppressors using small molecules329. Besides metallo and serine
proteases, cysteine proteases such as cathepsins are also known to play a causal role in tumor progression and metastasis. They are involved in invasion either by directly cleaving the ECM components or by indirectly activating other proteases such as pro-uPA and MMPs172. Hence, cystatins are likely to regulate tumor invasion and metastasis by inhibiting cysteine cathepsins.
Besides the cathepsin inhibitory property of CSTA, its involvement in apoptosis and differentiation makes it an interesting point of investigation for exploring novel avenues in cystatins-based treatment strategy330.
During tumor progression, polarized epithelial cells undergo multiple biochemical changes resulting in a mesenchymal cell phenotype with enhanced migratory potential. This process is called EMT. Members of the cystatin superfamily were proposed to impede EMT